Skip to content
[ 04 ]Work

Proof, not promises.

Most of our work sits under NDA, so what follows is anonymised — the shape of the problem, the layer it lived in, and what moved. Detailed case studies are available on request.

Sectors
Finance · Consumer · Health
Detail
Available under NDA
Reported
Measured, not estimated
[04Engagements]
01Financial ServicesLayers 01–03

Retrieval that survives an auditor

The situation

A research desk had a document assistant that answered fluently and cited badly. Confidence was high, grounding was not, and compliance had stopped the rollout.

What we did

Rebuilt retrieval as hybrid search with a cross-encoder reranker, rewrote chunking around document structure, and put a grounding evaluation in CI so a regression fails the build.

Answer groundedness
0×
Unsupported claims
0%
Compliance review
Passed
02Consumer PlatformLayers 06–09

The bill was in the seventh layer

The situation

Inference spend was growing faster than usage. The product team had optimised prompts twice and run out of surface-level ideas.

What we did

Moved the hot path to a distilled open model, applied 8-bit quantisation and continuous batching, and routed only the hard tail to a frontier API.

Cost per request
0%
Throughput
0×
Quality parity
Held
03HealthcareLayers 04–10

An assistant that never leaves the building

The situation

Clinical data could not cross the network boundary, which ruled out every hosted option the team had evaluated.

What we did

Designed and deployed an on-prem stack on existing GPU capacity — fine-tuned small models, private serving layer, full audit trail — sized against real utilisation rather than a vendor's quote.

Data egress
0
Idle GPU reclaimed
0%
Deployment
On-prem
[Under NDA]

The rest is a conversation.

We don't publish client names or screenshots. If you want the specifics — architecture diagrams, benchmark methodology, what went wrong on the way — ask, and we'll walk you through it.

Next step

Ask for the case study.

Tell us which of these looks closest to your problem and we'll bring the detailed version to the first call.

Case studies shared under mutual NDA.